European Spallation Source RF Systems Dave McGinnis RF Group Leader ESS Accelerator Division SLHiPP-1 Meeting 9-December-2011

Overview 5 MW proton linac – Pulse Length = 2.9 mS – Pulse Rate = 14 Hz – Beam Current = 50 mA – Energy = 2.5 GeV

ESS RF System Overview

System Bandwidth Dominated by large beam loading of 50 mA Spokes: – R/Q = 500 Ohms, V = 5.7 MV – Q L = 240,000, Bandwidth = 1500 Hz Medium Beta: – R/Q = 300 Ohms, V = 11.3 MV – Q L = 800,000, Bandwidth = 900 Hz High Beta: – R/Q = 470 Ohms, V = 17.3 MV – Q L = 750,000, Bandwidth = 940 Hz

De-Tuning Lorentz detuning – Max Gradient = 18 MV/meter – K L ~ 1.25 Hz/ (MV/m) 2 – Detuning ~ 400 Hz => 75 degrees – Time constant ~ 1mS Pulse length ~ 3 mS Cannot offset by a static de-tune Piezo-compensation looks to be necessary (unlike SNS) – Or else pay for it with RF power!!! Micro-phonics ~ 10 Hz => 2.0 degrees – Active damping by piezo-tuners does not seem necessary

RF Regulation Since Linacs are single pass, no overhead required for instabilities like in rings The majority of RF regulation can be compensated by adaptive feed-forward – Dynamic Lorentz detuning compensated by piezo tuners – Modulator droop and ripple are consistent pulse to pulse – Beam current droop and ripple are consistent pulse to pulse (especially H+ sources).

System Overhead Is required for pulse to pulse variations Required for beam startup – How much can the beam current be changed in between a single pulse interval and still accelerate the beam on the next pulse – This requirement will dominate the overhead requirements – ESS is currently working with a 25% overhead (SNS experience) – 5% for loss in distribution

Cavity Coupling Unlike electron linacs, the gradient profile along the linac is not flat. How is the coupling set? – At maximum gradient? – Minimal reflected power for a cavity family? Can we set coupling cavity-to-cavity? – Adjustable couplers not an option – Over-couple the cavity couplers and use Custom iris couplers in the waveguides Or stub tuning in the waveguides

Forward and Reflected Cavity Power during beam pulse Coupling set at optimum at Maximum gradient

Number of Power Sources Per Cavity At high energy, the issue becomes cost. Consider – Two 1.0 MW Klystron + two Modulators ~280 k€ /klystron + ~570 k€ /modulator 1700 k€ – Two 1.0 MW Klystron + one Modulator ~280 k€ /klystron + ~800 k€ /modulator 1360 k€ – One 2.0 MW klystron + one Modulator ~330 k€ /klystron + ~800 k€ /modulator 1130 k€ Savings = 230 k€ ~ 20% Neglects the cost of extra distribution or vector modulators

Number of Power Sources Per Cavity At low energy (beta), the question becomes cavity to cavity variations for vector regulation Cavity to cavity variations – Lorentz detuning variations and control (-> 70 degrees over three time constants) – Coupling variations – Field flatness Most likely would need fast vector modulation – About the same cost of a klystron? – Bandwidth limitations? – Power handling? – Efficiency? Long lead time for klystron procurement – klystron procurements would begin before vector modulation development can be completed For the Baseline – ESS will choose one power source per cavity

One Modulator Per Klystron Limited space for assembly and repair

Two Klystrons per Modulator

Modulator Cost (Carlos Martins) Capacitor charger power supplies: – 30% Capacitor banks: – 5% Solid state switch assembly(ies): – 15% Transformers (if existent): – 15% All other ancillaries – 10% Assembling and testing work + overheads: – 25%

Doubling the Power of a Modulator (Carlos Martins) Capacitor charger power supplies: – The rated power doubles. – In many topologies several identical modules are used in parallel. – In this case we should double the number of modules. – We can then consider that the cost of capacitor charger power supplies doubles; Capacitor banks: – The stored energy will double. – Indeed, since the current is doubled the capacitance value needs to be doubled for the same tolerated voltage droop. – The cost of capacitor banks will then double; Solid state switch assembly(ies): – The peak current and RMS current will double. – Depending on the topologies and switch technology adopted, this might have a little impact on the size and cost of the switch assemblies or might have an impact corresponding to a factor of 2. – Let’s suppose a factor of 1.5 in average; Transformers (if existent): – The fact that the peak power is roughly the double, the size of the transformer will be higher but not a factor of 2. – Assume about 30% in extra volume. – We can then consider that the price is multiplied by 1.3; All other ancillaries – cabinets, wiring, control system, HV cables, mechanical work, cooling circuit, electrical distribution components, etc. will be more or less the same. Assembling and testing work + overheads: – Will be about the same. Doubling the power of a modulator increases the cost by 1.45x

Modulator Requirements 109 modulators with one modulator for every two klystrons 3.46 mS pulse flat-top at a rate of 14 Hz 120kV and 40 Amperes at flat-top Cost range – 1.2 M€ per modulator – CERN Modulator = 0.7 M€ for 2.5mS flat-top at 20Hz with 120kV and 20A – 1.16x for longer pulse length, 1.45x for higher current Production rate 2 modulators per month for 4.5 years (Sept 2013 – March 2018) The Major Risk is Schedule Risk

Modulator Strategy Few number of vendors each with their own unique topology – For example, CERN modulator: 4 different vendors, 4 completely different topologies The only way to avoid minimize schedule risk is to have multiple vendors building the same modulator design. ESS must “own” the modulator design – key components of the modulator cannot be proprietary – vendors build-to-print. Multiple vendors can: – build complete modulators – Or build modulator parts with some vendors assembling complete modulators

The Baseline Design The Baseline Design Will Use Multiple Resonant Sub- Convertor Design – Advantages Open source topology All electronic active devices are at a medium-voltage level Semiconductor switches and drivers are of standard commercial types No demagnetization circuits are needed. The flat-top voltage (droop) is regulated in closed loop In case of klystron arcing, the resonant circuits will be automatically de-Q’d The topology and the mechanical layout are entirely modular. – Disadvantages Construction of the high frequency transformers can be challenging H-bridges handle a significant amount of reactive power Soft-switching of the IGBT’s in all operating points might be complex.

Resonant Modulator Concept

Modulator Backup Design The backup design will use Bouncer modulator topology. – Advantages Open source topology The power circuit is simple and reliable. All electronic active devices are at a medium-voltage level Voltage ripple on the flat-top is small Solid, reliable topology for long pulses Large experience at other laboratories. – Disadvantages Large pulse transformers and LC resonant bouncer volume for long pulses Slow rise and fall times Reverse voltage on the klystron to demagnetize the pulse transformer limits the duty cycle. Prototypes – The CERN 704 MHz test stand will use this topology and will be a pre- prototype for ESS.

Bouncer Modulator Concept

CERN Modulator Specifications

352 MHz Spoke Power Spoke Power – power sources – at 352 MHz – peak power capability of 370 kW What type of power source – Solid state – IOT – Triode/tetrode – Klystron What type of modulator – How many power sources per modulator (8?)

Sustainable Energy Concept Can we recover (re-use) the energy deposited in the collector of the klystrons? Renewable Carbon dioxide: -120,000 ton/y Responsible Carbon dioxide: -30,000 ton/y Recyclable Carbon dioxide: -15,000 ton/y

Summary The klystron modulator system – is likely to be the most costly accelerator component – And will have significant schedule risk – We propose an open source design and invite laboratory/university/industrial collaborators to participate in a consortium to develop the design We need a solution for the 352 MHz spoke cavity RF Vendorama – Lund, Sweden - February 2012